Search papers, labs, and topics across Lattice.
4
0
8
Forget slow, unconstrained diffusion models: Sketch2Colab distills a diffusion prior into a fast, controllable rectified flow for generating coordinated multi-human motion from sketches.
Forget expensive data generation and unstable PINNs: this method trains neural PDE solvers with cheap, noisy Monte Carlo estimates, achieving up to 8.75x improvement in L2 error.
By recasting the Hamilton-Jacobi-Bellman equation as a tractable Monte Carlo estimation, this work stabilizes physics-informed RL and unlocks its potential for high-dimensional control tasks.
LLMs can now plan robot manipulation of clutter for navigation in previously inaccessible environments, enabling zero-shot generalization in interactive object placement tasks.
